Faster, better, cheaper is the mantra of IT, except when it isn't. Most technology gets faster, better, cheaper over time, and unified communications (UC) is no exception. But the more exciting
Enabling UC throughout the network
Looking at UC mobility growth projections
Reducing operations expense with SDN in your WAN
Extending SDN to the enterprise campus network
Communications have always been central to work, but over the past few years, communications and collaboration have impacted and driven how and where work gets done.
Simply put, the UC revolution is about how work changes.
Throughout this disruption, the underlying networks got faster, better, cheaper -- especially wireless networks and WANs. Modern enterprise networks are faster, better, cheaper, too, despite the continued use of relatively mature technologies (100BASE-T; LAN switching; Power over Ethernet, or PoE; virtual local area networks, or VLANs; TCP/IP, etc.).
Now the networks are queued up for a big change, and the UC house is about to get new carpeting. This is more than faster, better, cheaper with existing technology. It's time to prepare for software-defined networking (SDN).
Software-defined networks to benefit 'needy' UC applications
Software-defined networks are the newest old idea in networking. The simplest description is that these networks separate the control and data functions. Most current network devices, like switches, are responsible for making routing decisions (the control function), as well as the actual forwarding of packets (the data function). There are benefits to separating these functions and centralizing network control.
UC is a killer app for end-to-end software-defined networks because it is dynamic and thirsty for responsive bandwidth.
Sound familiar? It should, for two reasons. First, the separation and virtualization of components already happened to servers in the data center. Server virtualization transitioned physical server components like processors and storage into abstract services allowing administrators to dynamically allocate resources as needed. Virtualization made the need to physically rebuild servers to expand or move resources an archaic practice.
It should also be familiar because software-defined networks are not the first attempt at creating a separate control function. The most popular example is modern Wi-Fi networks, as they frequently utilize separate controllers with multiple distributed radios (data). One could argue that X.25 and asynchronous transfer mode, or ATM, with their permanent virtual circuits, also used separate control and data planes.
Software-defined networks and UC go together like a horse and carriage. That's because over the past decade, communications transitioned from platform to application -- and a needy application at that. UC conversations are highly sensitive to network performance, plus users expect to be able to move conversations across devices mid-syllable. Few applications have such dynamic network requirements.
Server virtualization makes it relatively simple to move server applications across physical devices. The problem is that doing this conflicts with the localized learning that takes place within network switches. Packet switching is a Layer 2 technology -- faster than actual Layer 3 (network layer) routing -- but more limited. Moving applications, as with VMware vMotion, wreaks havoc on switch tables. SDN can facilitate the movement of server resources by centralizing updates to the network tables. Software-defined networks make sense in the data center, and UC applications make them attractive enterprise-wide.
The goal for UC is fast network provisioning as usage requires
It is important to remember that software-defined networks are a concept, not a specific technology. They involve controllers and application program interfaces (APIs) that can interface with applications. The goal is that a UC solution could seamlessly communicate, resulting in the dynamic provisioning of networks as usage dictates.
SIP meets SDN
How Juniper and Sonus Networks are partnering to to fix UC availability
Tightly integrating UC with the network could ensure that bandwidth and resources are properly provisioned. Unified communications is a killer app for end-to-end software-defined networks because it is dynamic and thirsty for responsive bandwidth.
For example, with the increased use of bring your own device (BYOD), an HD video conference can erupt anywhere without notice. When you combine this with the motivations for developing a software-defined network -- centralized control, automatic and dynamic provisioning, dynamic movement of virtual machines, improved analytics and improved security, it seems unlikely that these networks will be contained to data centers.
The most popular SDN protocols include OpenFlow, shortest path bridging (SPB), TRILL, and Open Shortest Path First (OSPF). At this time, a single winner or standard is unlikely. There are also multiple frameworks including OpenDayLight and OpenStack, and vendor-specific solutions including Juniper's Contrail V, Nuage Networks' Virtualized Services Platform (VSP), Cisco's onePK, and VMware's NSX. The good news is that SDN doesn't necessarily mean having to buy all new network switches. OpenFlow, for example, works with many existing switches.
UC vendors making the SDN connection
The vendors that offer the triple threat of networking, UC and UC endpoints (hardware or software) will likely be the first to deliver end-to-end UC SDN solutions. These vendors include Avaya, Cisco, Alcatel-Lucent and NEC (though HP did a notable demonstration of SDN functionality with Microsoft Lync using an unreleased API).
Software defined networks are faster, better, cheaper, but they represent far more than an incremental upgrade and could very well radically impact how UC gets delivered.
This was first published in January 2014